File size: 7,196 Bytes
8b7a3d1 400839c db0cd98 8b7a3d1 bd420fb 400839c a3f0757 8b7a3d1 c0a7c3c 8b7a3d1 a3f0757 8b7a3d1 a3f0757 8b7a3d1 a3f0757 8b7a3d1 07a02e5 a3f0757 07a02e5 a3f0757 07a02e5 a3f0757 07a02e5 a3f0757 07a02e5 a3f0757 07a02e5 a3f0757 07a02e5 a3f0757 8b7a3d1 c0a7c3c 07a02e5 8b7a3d1 bd420fb 07a02e5 8b7a3d1 400839c bd420fb 400839c 8b7a3d1 87968ab 400839c 8b7a3d1 db0cd98 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 |
#!/usr/bin/env python
from __future__ import annotations
import os
import gradio as gr
from app_system_monitor import create_monitor_demo
from constants import UploadTarget
from inference import InferencePipeline
from trainer import Trainer
def create_training_demo(trainer: Trainer,
pipe: InferencePipeline | None = None,
disable_training: bool = False) -> gr.Blocks:
def read_log() -> str:
with open(trainer.log_file) as f:
lines = f.readlines()
return ''.join(lines[-10:])
hf_token = os.getenv('HF_TOKEN')
with gr.Blocks() as demo:
with gr.Row():
with gr.Column():
with gr.Box():
gr.Markdown('Training Data')
training_video = gr.File(label='Training video')
training_prompt = gr.Textbox(
label='Training prompt',
max_lines=1,
placeholder='A man is surfing')
gr.Markdown('''
- Upload a video and write a `Training Prompt` that describes the video.
''')
with gr.Column():
with gr.Box():
gr.Markdown('Training Parameters')
with gr.Row():
base_model = gr.Text(
label='Base Model',
value='CompVis/stable-diffusion-v1-4',
max_lines=1)
resolution = gr.Dropdown(choices=['512', '768'],
value='512',
label='Resolution',
visible=False)
input_token = gr.Text(label='Hugging Face Write Token',
placeholder='',
visible=False if hf_token else True)
with gr.Accordion('Advanced settings', open=False):
num_training_steps = gr.Number(
label='Number of Training Steps',
value=300,
precision=0)
learning_rate = gr.Number(label='Learning Rate',
value=0.000035)
gradient_accumulation = gr.Number(
label='Number of Gradient Accumulation',
value=1,
precision=0)
seed = gr.Slider(label='Seed',
minimum=0,
maximum=100000,
step=1,
randomize=True,
value=0)
fp16 = gr.Checkbox(label='FP16', value=True)
use_8bit_adam = gr.Checkbox(label='Use 8bit Adam',
value=False)
checkpointing_steps = gr.Number(
label='Checkpointing Steps',
value=1000,
precision=0)
validation_epochs = gr.Number(
label='Validation Epochs', value=100, precision=0)
gr.Markdown('''
- The base model must be a Stable Diffusion model compatible with [diffusers](https://github.com/huggingface/diffusers) library.
- Expected time to train a model for 300 steps: ~20 minutes with T4
- You can check the training status by pressing the "Open logs" button if you are running this on your Space.
''')
with gr.Row():
with gr.Column():
gr.Markdown('Output Model')
output_model_name = gr.Text(label='Name of your model',
placeholder='The surfer man',
max_lines=1)
validation_prompt = gr.Text(
label='Validation Prompt',
placeholder=
'prompt to test the model, e.g: a dog is surfing')
with gr.Column():
gr.Markdown('Upload Settings')
with gr.Row():
upload_to_hub = gr.Checkbox(label='Upload model to Hub',
value=True)
use_private_repo = gr.Checkbox(label='Private', value=True)
delete_existing_repo = gr.Checkbox(
label='Delete existing repo of the same name',
value=False)
upload_to = gr.Radio(
label='Upload to',
choices=[_.value for _ in UploadTarget],
value=UploadTarget.MODEL_LIBRARY.value)
remove_gpu_after_training = gr.Checkbox(
label='Remove GPU after training',
value=False,
interactive=bool(os.getenv('SPACE_ID')),
visible=False)
run_button = gr.Button('Start Training',
interactive=not disable_training)
with gr.Box():
gr.Text(label='Log',
value=read_log,
lines=10,
max_lines=10,
every=1)
if not disable_training and not os.getenv(
'DISABLE_SYSTEM_MONITOR'):
with gr.Accordion(label='System info', open=False):
create_monitor_demo()
if pipe is not None:
run_button.click(fn=pipe.clear)
run_button.click(fn=trainer.run,
inputs=[
training_video,
training_prompt,
output_model_name,
delete_existing_repo,
validation_prompt,
base_model,
resolution,
num_training_steps,
learning_rate,
gradient_accumulation,
seed,
fp16,
use_8bit_adam,
checkpointing_steps,
validation_epochs,
upload_to_hub,
use_private_repo,
delete_existing_repo,
upload_to,
remove_gpu_after_training,
input_token,
])
return demo
if __name__ == '__main__':
hf_token = os.getenv('HF_TOKEN')
trainer = Trainer(hf_token)
demo = create_training_demo(trainer)
demo.queue(api_open=False, max_size=1).launch()
|